Bittensor (TAO)

New launch from the @YumaGroup subnet accelerator: @SubQueryNetwork and @graphprotocol already power 50k+ projects Hermes (SN82) brings agentic AI built on Bittensor $TAO to all of them https://t.co/tFFxYhguGi

NEW from Yuma’s Subnet Accelerator: Hermes (SN82) Crypto apps and trading tools rely on blockchain data, but most apps rely on slow and expensive AI systems to get the answers they need. @HermesSubnet is tackling the issue, using Bittensor $TAO to train AI agents to pull and respond with onchain data like token prices, wallet history, or governance actions – making it easy for end users to query with natural language.

$TAO #SN82 - It seems like subnet number 82 is not unknown anymore. A new @YumaGroup team entered the chat! And it also seems like they arrived well prepared, looking at how complete their first Discord announcement is. Welcome to the arena, @SubQueryNetwork. -- "Hermes is a Bittensor subnet built to act as the messenger between AI agents and blockchain data (much like the god it’s named after). It connects AI agents directly to blockchain data via GraphQL, the cornerstone of Web3 data access. A new wave of AI is transforming industries, but Web3 risks missing out. The current approach of using large, general‑purpose LLMs for GraphQL agents is highly costly, with poor latency and efficiency, stifling real innovation and production‑ready use cases. Hermes addresses these challenges by training, testing, and rewarding AI systems that can understand and query on‑demand Web3 data — turning complex on‑chain information into simple, intelligent answers." TEAM The Hermes subnet is built by the SubQuery team. • Ian He — CTO (@ian_he) • Jacob Zhou — Lead Engineer (@jacobzo) • Cyrbuzz — Fullstack Engineer • Luke — Subnet Engineer • Brittany Seales — Head of Marketing (@Brittany Seales) • Marta Adamczyk — Head of BD (@deverka) LINKS Website: https://t.co/hxfxcUheWR GitHub: https://t.co/Q4KwmqJ2pY Litepaper: https://t.co/FkjvscDim0

JSTprove JUST WENT LIVE ON BITTENSOR SN2 DSperse $TAO Almost nobody is talking about what this actually means. Might be the most important infrastructure development in decentralized AI right now. THE PROBLEM AI's dirty secret: You have no idea if the model actually did what it claims. When an agent says it analyzed your portfolio, ran a diagnostic, or executed a trade on specific parameters, you're trusting a black box. It could lie. Use different weights. Hallucinate. Zero cryptographic proof the inference happened as described. Fine for memes. Catastrophic for money, medicine, or autonomous systems. The entire AI agent economy runs on TRUST ME BRO! PROOF OF INFERENCE @Inference_Labs built the fix at protocol level. JSTprove: ZK toolkit proving a specific model produced a specific output on specific inputs, verifiable by anyone, without exposing weights, data, or inputs. Just went live after testnet. Scaling to thousands of proofs/week. Unlocks verifiable inference for complex models previously impossible with ZK. Previous zkML: Needed crypto PhDs, couldn't handle real models, too slow. JSTprove: Abstracts complexity, 65-76% faster proving, <1GB memory via model slicing + DSperse. Paper (Oct 2025): Full ONNX → ZK pipeline. Simple CLI: compile, witness, prove, verify. TRUTHTENSOR IN PRODUCTION TruthTensor launched 40 days ago: • 1M+ agent fine-tunes • 500K+ active trading agents • 500K+ builders Proof of Portfolio: ZK-verified PnL without leaking strategies (audited ZKVIN/Zellic). 283M+ proofs processed late 2025, scaling fast. Cysic ASIC partnership: Cheaper/faster verifiable inference. This isn't research. It's shipping at scale. AGENT ECONOMY Every serious agent builder hits the trust wall: • Trading: Prove strategy was followed? • Medical: Prove approved protocols used? • Robotics: Prove safety parameters respected? Until now: Audits, reputation, hope. Doesn't scale. Proof of Inference = mathematical receipt. Anyone verifies independently. No faked logs. No trusted attestors. Pure crypto truth. $TAO CONNECTION Subnet 2 is core infrastructure. Every high-stakes subnet needs this: • Ridges proves coding agents used claimed models • Synth proves signals from verified compute • Any finance/healthcare/logistics agent gains trust layer Value accrues to $TAO: Verifiable inference becomes primitive consumed network-wide. AI agents are coming for critical decisions finance, healthcare, transport, security. That future can't run blind. JSTprove + SN2 deliver the cryptographic foundation: mathematical certainty over reputation. Few are building this. Even fewer see why. Code ships. Proofs generate. Market sleeps. Question: How many see it before it's priced in? $TAO Sources 🔗 https://t.co/HJbMZL1Kkj https://t.co/OaBllUup6B https://t.co/KzQd4M3T9g

Ah, this is nice Alex ... TAO Flow for all subnets -- at a glance. https://t.co/QJ6ovxUe0C

$TAO #SN52 - Erratum, the Binance funded subnet, @TensorplexLabs, is dead for good 🥲 https://t.co/DGBDILY8hf